Benchmarking Dual-Polarization Sentinel-1 RTC Against Complex-SLC InSAR Coherence for Open-Pit Mining Disturbance and Tonnage Reconciliation: Antamina (Peru) and Mount Milligan (Canada)
Year: 2026
Type: preprint
Abstract: Global greenhouse gas (GHG) inventories such as Climate TRACE estimate open-pit mining production from Synthetic Aperture Radar (SAR) disturbance proxies, yet full-frame Sentinel-1 Single Look Complex (SLC) interferometry requires staging ~1.2 TB of raw bursts for a two-mine, three-year audit. This study tests whether cloud-streamed 10 m dual-polarization Radiometrically Terrain-Corrected (RTC) backscatter stability can reproduce complex-SLC interferometric phase coherence for Normalized Difference Activity Index (NDAI) and Hue-Saturation-Value (HSV) mine tonnage reconciliation, paired with a burst-level complex-SLC cloud pipeline. First, a strictly non-circular, co-orbital, co-temporal validation from December 2019 to November 2020 (76 twelve-day RTC pairs) is conducted against NASA/JPL GAMMA complex-SLC 12-day phase coherence (COH12) across 5,274 native 90 m cells at Antamina, Peru (Orbit 069D; 4,300 m ASL) and Mount Milligan, British Columbia, Canada (Orbits 064A/137A; 1,100 m ASL). With zero spatial fitting parameters, dry-season RTC coherence correlates strongly with GAMMA complex-SLC coherence inside both footprints (r = 0.677 at Antamina; r = 0.690 and NDAI r = 0.728 at Mount Milligan) and discriminates active mining with ROC-AUC of 0.811 and 0.879. Because positive Rayleigh intensity correlation approaches pi/4 (~0.785) rather than zero under complete phase decorrelation, single-parameter power scaling compresses coherence to 0.52-0.84; percentile-stretched HSV segmentation is invariant to this floor (overall accuracy 77.0% and 82.8%), while out-of-sample cross-mine floor subtraction (RTC-2P) restores linear NDAI dynamic range (RMSE = 0.214-0.219). Deploying the workflow across 171 twelve-day Sentinel-1A/1C pairs in 2023-2025 removes all 54 winter snow pairs in British Columbia and 31 Andean storm pairs (rho_stable < 0.80), while single- and dual-burst SLC manifests cut raw archive volume by 96.1% (from 1,199.9 GB to 46.50 GB; 253 NASA ASF HyP3 credits). Steady-state 2023-2024 total rock moved per active area converged at 30.23-32.55 t/m2 (Antamina) and 25.85-27.43 t/m2 (Mount Milligan), whereas 2025 capital pushbacks and tailings beaches (53.8-65.9% of active pixels) inflated single-year area by 26.0% and 62.2%, showing why pit/dump vs. tailings sub-zoning and multi-year smoothing (+2.2% and +6.8% cumulative error) are needed. Accompanying archive (Supplementary_Data_and_Code_Archive.zip) contains all Python processing scripts, ESA SNAP XML interferometric graphs, KML/GeoJSON mine boundaries, 10-pair and 171-pair NASA ASF HyP3 burst manifests, statistical benchmark and reconciliation CSV/JSON tables, and 10 m UTM GeoTIFF rasters.
Author: Bahukhandi, Vijay
Language: English
Subfield: Geophysics
Field: Earth and Planetary Sciences
Domain: Physical Sciences
SDG: Climate action
Open Access status: green
OpenAlex ID: W7225452386